1 FUNCTIONS

2 FISH ORDINATION

This is an up-dated fish file with only the “final” fish multivariate analyses for the Flagstone MS.

2.1 Import and manipulate data

For data processing:

Corrected some spp codes
Make a ‘group’ variable for lumping taxa/species
Combined 2015 and 2016-2019 data
- 2015 data are not necessarily used in analyses. See below.
- 2015 data lack visibility estimates, so are usually deleted.
No depth data for the 2015 data.
- I took the average of the quadrats.
- I assigned transects with an average depth of >20 ft to the 10-m group.

NOTE: Transects with visibility < 2 m were deleted from the analysis.

NOTE: These analyses include 2015.

DATA TRANSFORM: none

I set the data transform to ‘none’ because there wasn’t a lot of variation in spp density among groups. Some but not a lot. Again, this emphasizes the more abudnant species, which is, I think, ok.

NOTES

Some rockfishes were recorded in the ‘adult’ category (eg SECA instead of SECAy) but were small.
Set minimum size for rockfishes to 10 cm. Fish <= to 10 cm were classified as YOY.
Maintained their original designation in SPECIES column. New designation in the ‘group’ column.

Because there were a lot of zeros for some species, we lumped some species or taxa into higher taxonomic categories. Species higher grouping used in the analyses are:

##    Spp_Code Spp_Group            Common_Name
## 1  BAITBALL      BAIT  bait-sardines-anchovy
## 4    SEBYTy    SEBYTy   yellowtail-black yoy
## 5      SEME      SEME         black rockfish
## 8      SEMY      SEMY          blue rockfish
## 12     SCMA      SCMA                cabezon
## 14    SEPIy     SEPIy    canary rockfish YOY
## 15     SENE      SENE         china rockfish
## 17     SECA      SECA        copper rockfish
## 20     HEXA      HEXA        Hexagrammos spp
## 23     CLUP      BAIT                Herring
## 25     HEDE      HEXA         Kelp Greenling
## 26     BRFR      EMBI         Kelp Surfperch
## 29     OPEL      OPEL                Lingcod
## 34     ENMO      BAIT       Northern Anchovy
## 38     SASA      BAIT        pacific sardine
## 39     OXPI      HEXA      Painted Greenling
## 41     RHVA      EMBI             Pile perch
## 45     HEHE      HEXA         Red Irish Lord
## 47     HELA      HEXA         Rock Greenling
## 49     RYOY      RYOY           rockfish_yoy
## 55     CYAG      EMBI       Shiner Surfperch
## 60     EMLA      EMBI      Striped Surfperch
## 61     EMBI      EMBI            Surfperches
## 63     AUFL      AUFL              Tubesnout
## 66     HEST      HEXA Whitespotted greenling
## 69    SECAy     SECAy    Copper rockfish YOY
## 70    SEMEy     SEMEy     black rockfish YOY

This list may have been sub-setted again. See below.

2.1.1 Year x Site x Depth

No text here. Just calculating averages by year x site x depth and outputting data.

2.1.2 Year x Site

No text here. Just calculating averages by year x site and outputting data.

2.2 Constrained Ordination: Year + Site + Depth (zone)

The ordinations here use transect level information in constrained ordinations. The PerMANOVA uses Year, Site, and Depth as Factors. The ordination is and RDA-type approach to Canonical Analysis of Principal Coordinates (not exactly the same) using the ‘capscale’ package. I present centroids and se, not transects in the figures for clarity.

Note: This analysis does NOT include rockfish YOY. We deleted YOY from the analyses because they were highly variable and often appeared in only one year. Univariate plots do show YOY.

DATA TRANSFORM: none

Taxa Included: OPEL, HEXA, EMBI, AUFL, BAIT, SECA, SCMA, SENE, SEME

2.2.1 Permanova

In the PerMANVOA everything is significant. I think this is OK. Things are messy. I don’t actually think it is necessary for our paper.

The associated ordination is more clear. This ordination uses site x year x depth as groups and transects as replicates. I have, however, calculated the centroids for plotting.

FYI removing BAIT doesn’t do much.

Bray-Curtis transform.

## Permutation: free
## Number of permutations: 999
## 
## Terms added sequentially (first to last)
## 
##                 Df SumsOfSqs MeanSqs F.Model      R2 Pr(>F)    
## SITE             4    14.493  3.6234  24.959 0.11590  0.001 ***
## ZONE             1     8.465  8.4649  58.309 0.06769  0.001 ***
## YEAR             1     3.378  3.3776  23.266 0.02701  0.001 ***
## SITE:ZONE        4     4.285  1.0712   7.379 0.03426  0.001 ***
## SITE:YEAR        4     5.641  1.4103   9.715 0.04511  0.001 ***
## ZONE:YEAR        1     2.365  2.3647  16.289 0.01891  0.001 ***
## SITE:ZONE:YEAR   4     2.225  0.5563   3.832 0.01780  0.001 ***
## Residuals      580    84.200  0.1452         0.67332           
## Total          599   125.052                 1.00000           
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

2.2.2 capscale - site x year x depth

To match PerMANOVA. RDA type analysis with permutations. This analysis has the same Factor groupings as the PerMANOVA.

Ordination plot for CAP based on site x year x depth. Error bars are +/- 1.0 s.e. Lower-right pane zooms in on the species scores.

Ordination plot for CAP based on site x year x depth. Error bars are +/- 1.0 s.e. BAIT is not shown but is to the lower right outside the current axes limits.

Same data replotted on one pane. We can see some clear site differences:

Cape Alava and Cape Johnson do not overlap
Tatoosh and Neah Bay are intermediate between the former two
Destruction is all over the place

3 INVERT ORDINATION

Groupings used in the Invert Ordination were:

bivalve, blood_star, brood_sea_star, Cal_cuc, chiton, crabs, green_urchin, hermit_crabs, kelp_crab, large_anemone, large_barnacle, large_nudibranch, large_sea_star, leather_star, med_nudibranch, med_sea_star, orange_cucumber, P_ochraceous, purple_urchin, red_urchin, sea_cucumber, sea_star_YOY, shelled_gastropod, shelled_mollusk, small_anemone, sponge, tunicate

The invert ordination was pretty clear. Tatoosh, for exmaple, was characterized by large numbers of all urchins. Destruction, Alava, and Tatoosh all had some Pisaster and leather stars.

Ordination plot for CAP based on site x year x depth. Error bars are +/- 1.0 s.e.

4 HABITAT VARIABLES

I used PCA for the habitat variable ordinations instead of a constrained ordination or nMDS in order to do data reduction and produce variables to include in future ordination or other analyzes (see below). Constrained ordinations are not really data reductions and nMDS doesn’t really produces usable axes. Thus PCA seemed the best approach.

PCA the various habitat variables to reduces dimensions
Do separately for each variable type (substrate, UPC)
Initial analysis by depth zone to maintain the most replicates
- not necessarily interested in depth for the analyses
Did NOT include kelp to maintain a separation between biotic and abiotic habitat

4.1 Substrate PCA - YEAR x SITE x DEPTH means

PC1 mostly distinguishes between bedrock and boulder. Boulder areas had higher relief (>2m) and higher relief diversity. PC2 tends to distinguish between the two mid-complexity categories.

While there is some variation among years within sites (spread), it isn’t bad. Mostly, there are some obvious differences among sites with regard to habitat.

This might be a good initial ordination figure just describing the sites and showing that the physcial habitat differs among sites. Depths to a lesser extent.

Tatoosh and Destruction Island have more bedrock and higher relief than other areas.
Neah Bay is bascially intermediate for everything.
Cape Alava and Cape Johnson have more boulder and kind of intermediate relief.

Substrate PCA. Open circles are 5-m depth zone; closed circles 10-m depth zone.

4.2 KELP PCA

The kelp ordination isn’t great and we can’t really reduce the axes much more than the four axes in the original data. Also, given only four groups and three actual species, I think it is better to maintain the identity and use the original data not PCs in the following analyses.

You can see some depth pattern with deeper sites (filled circles) off to the upper left. This is hardly surprising.

4.3 UPC - PCA

There is some separation of points based on site:

UPC PCA analysis. Open circles are 5-m depth zone; closed circles 10-m depth zone.

5 FISH VS ABIOTIC HABITAT AND KELP

I combined the PCs from the habitat analysis with the kelp data. I then ran and RDA style constrained analysis.

This constrained habitat vs. fish analysis is ugly. In fact, it is non-significant! So while there are differences among sites x depth x year in fish abundance, they don’t seem related to habitat directly.

This actually makes sense when you compare the separate fish and habitat ordinations. For example, Cape Alava and Cape Johnson have different fish fauna, but similar habitat.

##  [1] "YEAR"   "SITE"   "ZONE"   "Subs_1" "Subs_2" "upc_1"  "upc_2"  "MACPYR"
##  [9] "NERLUE" "PTECAL" "OTHER"

## Call: capscale(formula = cap_data[, cap.fish] ~ Subs_1 + Subs_1 + upc_1
## + upc_2 + MACPYR + NERLUE + PTECAL + OTHER, data = cap_data)
## 
##                 Inertia Proportion Rank
## Total         1299.9421     1.0000     
## Constrained    175.1500     0.1347    7
## Unconstrained 1124.7921     0.8653    9
## Inertia is mean squared Euclidean distance 
## Species scores projected from '[' 'cap_data' '' 'cap.fish' 
## 
## Eigenvalues for constrained axes:
##   CAP1   CAP2   CAP3   CAP4   CAP5   CAP6   CAP7 
## 161.57  11.68   1.36   0.47   0.05   0.02   0.00 
## 
## Eigenvalues for unconstrained axes:
##   MDS1   MDS2   MDS3   MDS4   MDS5   MDS6   MDS7   MDS8   MDS9 
## 1078.7   39.3    4.9    1.2    0.3    0.2    0.1    0.1    0.0

## Permutation test for capscale under reduced model
## Permutation: free
## Number of permutations: 999
## 
## Model: capscale(formula = cap_data[, cap.fish] ~ Subs_1 + Subs_1 + upc_1 + upc_2 + MACPYR + NERLUE + PTECAL + OTHER, data = cap_data)
##          Df Variance      F Pr(>F)
## Model     7   175.15 0.6229  0.699
## Residual 28  1124.79

## INVERTS

## Call: capscale(formula = cap_data_inverts[, cap.inverts] ~ Subs_1 +
## Subs_1 + upc_1 + upc_2 + MACPYR + NERLUE + PTECAL + OTHER, data =
## cap_data_inverts)
## 
##               Inertia Proportion Rank
## Total          1.5497     1.0000     
## Constrained    0.6785     0.4378    7
## Unconstrained  0.8712     0.5622   27
## Inertia is mean squared Euclidean distance 
## Species scores projected from '[' 'cap_data_inverts' '' 'cap.inverts' 
## 
## Eigenvalues for constrained axes:
##   CAP1   CAP2   CAP3   CAP4   CAP5   CAP6   CAP7 
## 0.6299 0.0349 0.0063 0.0036 0.0023 0.0012 0.0004 
## 
## Eigenvalues for unconstrained axes:
##   MDS1   MDS2   MDS3   MDS4   MDS5   MDS6   MDS7   MDS8 
## 0.5158 0.2907 0.0253 0.0127 0.0091 0.0063 0.0056 0.0023 
## (Showing 8 of 27 unconstrained eigenvalues)

## Permutation test for capscale under reduced model
## Permutation: free
## Number of permutations: 999
## 
## Model: capscale(formula = cap_data_inverts[, cap.inverts] ~ Subs_1 + Subs_1 + upc_1 + upc_2 + MACPYR + NERLUE + PTECAL + OTHER, data = cap_data_inverts)
##          Df Variance      F Pr(>F)   
## Model     7  0.67852 3.4492  0.002 **
## Residual 31  0.87119                 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

The Invert vs Habitat relationship was significant. The main difference separated Tatoosh from other areas. Interestingly, * Urchins in general were both positively and negatively associated with Nero.
* On Axis 1 Nero, Ptero and urchins are all off to the right. * On Axis 2 the urchins are positive while the kelps are more negative.

The postive relationship between UPC_1 and urchins in general puts the urchins at sites with more brown algae and less red algae.
The positive relationship with Subs_1 on the x-axis suggests more urchins at areas with bedrock.
- The negative relationship with Subs_1 on the y-axis suggests the relationship is more complex
- Perhaps urchins like areas with both bedrock and boulder. Broad areas to eat, some crevices to hide.

Ordination of invertebrate density vs. habitat charcteristics. Note, the second pane zooms in on the central cluster of points and exclues some points seen in the first pane.

6 UNIVARIATE PLOTS

These plots include all the “groups” in the data. These taxa-groups are not all in the ordinations above. They are plotted here for referecne

6.1 FISH

Fish abundance by site and year.

6.1.1 YOY plots

Two heat plots for rockfish YOY. They are the same data, just ordered differently to emphasize either species or site.

The first heat plot shows clearly the different recruiment patterns among species with * Blacks and Black/YT recruiting heavily in 2016 * Yelloweye (SEPI) in 2018 * Copper/quill and unidentified RYOY in 2019.

NOTE: colors are scaled across rows.

Heat-map plot of YOY abundance

6.2 INVERTS